causality
Find the lost submarine
It is June 1968, and the USS Scorpion has vanished in the Atlantic with ninety-nine men aboard — last heard from near the Azores, presumed down somewhere in thousands of square miles of ocean two miles deep. The Navy's search budget is finite and winter is coming. Run the search as Bayesian inference: elicit scenarios from submariners, weight them into a prior over the seabed, overlay acoustic bearings of dubious provenance, and update the map after every unsuccessful sweep — because failing to find the boat in a square is itself evidence. Allocate ships to maximize probability of detection per day. Get it wrong and the wreck — and the answer to what killed the crew — stays lost; the Bayesian map found it within a few hundred yards.
Who this problem belongs to
The two figures whose methods fit it best, out of 65 in contention.
Blackwell spent years at RAND in the 1950s working the intersection of game theory, dynamic programming, and sequential decision-making under uncertainty, exactly the machinery Craven's team assembled for the Scorpion hunt. His Rao-Blackwell theorem is about extracting maximal information from noisy observations, and his work with Arrow and Girshick on Bayes sequential decision procedures gave operations research the language for updating a belief state after each unsuccessful action and choosing where to look next to maximize expected information gain. He was not on that ship search personally, but the RAND culture he helped define, treating search as a sequential Bayesian allocation problem rather than a coverage problem, is precisely the intellectual lineage the 1968 team drew on. Only the absence of a documented direct hand keeps this shy of the very top score.
Laplace is the reason updating belief after evidence is even a coherent operation. His inverse-probability memoirs of the 1770s-1810s worked out, by hand, how to combine a prior over an unknown cause with observed data to produce a posterior, the identical logical structure Craven's team used to combine submariner scenario elicitations with failed-sweep evidence. Laplace also pioneered treating astronomical observation errors probabilistically to pin down unseen bodies from partial, noisy sightings, a genuine ancestor of localizing an unseen object from indirect acoustic and drift evidence. He never touched sonar or grid search, and his tools were closed-form integrals rather than discretized probability grids over a seabed, so the translation to an operational search plan required two more centuries of engineering. The reasoning skeleton, though, is entirely his.
Fought here
In the mind map
The same ideas, as concepts rather than history — in John's ML knowledge map.
65 figures are scored on this problem. Draw it in a battle to see where you land.